utf8check2.c

#include <stdio.h>
#include <stdint.h>
#include <stddef.h>

#ifndef __cplusplus
#define true 1
#define false 0
#endif

int check_utf8(unsigned const char* data, size_t length) {
    int expect_extend = 0;
    unsigned const char* end = data + length;
    for(;data < end; data++) {
        //Two highest bits set, beginning of a multi-byte sequence.
        if(*data >= 0xC0) {
            uint8_t rest = (*data & ~0xC0);
            ptrdiff_t end_dist = end - data;
            if(!(rest & 0x20) && end_dist > 1) {
                if((*(++data) & 0xC0) != 0x80) return false;
            } else if(!(rest & 0x10) && end_dist > 2) {
                if((*((uint16_t*)(++data)) & 0xC0C0) != 0x8080) return false;
                ++data;
            } else if(!(rest & 0x8) && end_dist > 3) {
                if((*((uint32_t*)(data)) & 0xC0C0C000) != 0x80808000) return false;
                data += 3;
            } else {
                //Invalid multi-byte sequence
                return false;
            }
        } else if(*data > 0x7F) {
            return false;
        }
    }
    //Should not be expecting more chars at end of string
    return true;
}

#define CHECKS(X) do { if(check_utf8((unsigned char*) X, sizeof(X) - 1)) { printf("%s - yes\n", X); } else { printf("no\n"); } } while(0);
int main(int argc, char** argv) {
    CHECKS("normal");
    //two-byte char
    CHECKS("breve'd a: ă");
    //three-byte char
    CHECKS("airplane: ✈");
    //four-byte char
    CHECKS("parenb: 🄑");
    //char outside of seq
    unsigned char invala[] = {128,0};
    //unexpected ascii char in mb-seq
    unsigned char invalb[] = {226,65,0};
    //too short at end
    unsigned char invalc[] = {65,226,0};
    //mb-seq inside mb-seq
    unsigned char invald[] = {226,226,65,67,0};
    CHECKS(invala);
    CHECKS(invalb);
    CHECKS(invalc);
    CHECKS(invald);
    return 0;
}